Ten days ago The Wall Street Journal said it had found strong statistical evidence of widespread vote manipulation in Russia’s recent parliamentary elections.
I had heard about the stories circulating in the Russian blogosphere about statistical analyses that seemed to show strange patterns in the official results for the governing United Russia party, leading some amateur analysts to conclude that many polling stations manufactured votes for the party of Vladimir Putin and Dmitry Medvedev – even though United Russia scored much worse than in the previous elections, winning 49 percent compared to 64 percent in 2007. And some results look mighty unlikely, no statistics needed, such as the amazing 99-plus percent both in turnout and United Russia support in the republic of Chechnya.
But it was hard to judge the credibility of the analyses. Now that the Journal claims to have confirmed the validity of the statistics, I’m going to wade in, at the risk of breaking my own rule about taking “statistically valid” claims at face value. If I’ve put my foot in it, I hope someone out there will call me on it, not because I like being shown up as a fool but because this is potentially a very big story involving deep manipulation of an election in the world’s largest country.
And also because critics of the Putin-Medvedev regime took these allegations very seriously. Some marched in protests with banners reading “We are for normal distribution” – “normal” meaning the common, bell-shaped curve mapping the relation between two variables, in this case turnout and United Russia support. (See the photo on this page, from Норвежский Лесной’s Livejournal page.) The fact that the results showed a two-humped curve, with the second hump peaking at 100 percent turnout, was interpreted by some as evidence of an artificially high level of United Russia support.
Wall Street Journal reporters Gregory L. White and Rob Barry used a computer to mine the election data on the website of the Russian Central Election Commission and analyzed it in consultation with unnamed experts in election fraud and statistics. (A Russian physicist and amateur election analyst claims the Journal used his methodology without attribution, but this is a side issue.)
The Journal’s analyses and others posted by Russian bloggers reach these conclusions:
● Many precincts reported turnout well above the national average of 60 percent, and 7,000 reported turnout of 97 to 100 percent. United Russia did unexpectedly well (and other parties correspondingly poorly) in these precincts. The party was credited with 3-plus million votes in precincts with turnouts of over 97 percent. “Such extra ‘turnout’ could be accounted for by fraud such as ballot-box stuffing, election experts say,” White and Barry write.
● In precincts where United Russia got a high percentage of the votes, its share of the vote tended to be reported in round numbers (65, 70, 75 percent, etc., up to 100 percent). Again, this may indicate that United Russia votes were manufactured in those precincts and the totals were then rounded off, the Journal article says. (It remains a question why election officials would be so stupid as to produce nice round numbers.)
Critiques of these analyses (here and here) say the authors have done bad statistics. Turnout and voting patterns do not necessarily form nice normal distributions, they say, because voting, unlike say the distribution of IQ or height, is not a random process, and the variables are not independent (areas with very high turnout could happen to coincide with pockets of very high United Russia support, for instance). The apparent spikes at round number percentages may be related to precinct size, they say: obviously, a precinct with say, 10 voters (and such precincts exist in Russia), is more likely than a large precinct to produce “rounded” results, e.g. 50 percent for candidate A, 25 percent for candidate B, and so on.
My jury is still out on the charge that United Russia was credited with millions more votes than were actually cast for the party. It seems to me that the election results show beyond a doubt that voting patterns in certain regions are unusual, to say the least, calling into question the mandate won by Putin’s party. Chechnya again this time showed its truly extravagant admiration for Putin, but no one really expects fair voting there. But there were also patterns of extremely high turnout and high support for United Russia in Dagestan, Bashkortostan, Tatarstan, and Mordovia.
Even if statistical flaws are found in these analyses, and even if the anecdotal reports of vote-rigging are shown to be fakes, or too few to have affected the overall results, the official results themselves suggest that in some parts of Russia, people either are pressured to turn out and to vote a certain way, or that votes are cast improperly in large numbers. And that comes on top of the ruling party’s domination of television and the official harassment of opposition candidates (although that tactic seems to have been used less this time than in 2007, when the “liberal” opposition was virtually crushed out of existence).
S.G. Kuznetsov on eruditor.ru points out that in the 2010 U.K. elections the Conservative Party, like United Russia, got a higher proportion of the vote as turnout rose. But imagine the hoo-ha if hundreds of British voting districts recorded 100 percent turnout, regardless of which party won! Or if the Conservatives won 100 percent in any district, regardless of turnout!